基于蚁群算法的支持向量机在矿区地表沉降预测中的应用研究

发布时间：2018-03-03 14:05

本文选题：支持向量机　切入点：蚁群算法　出处：《江西理工大学》2015年硕士论文　论文类型：学位论文

【摘要】：矿区地表沉降是一个缓慢且长期的过程,随着沉降量的累积,沉降总量超过地表所能承载的最大阈值,就会发生滑坡等重大灾害,为了避免灾害的发生,长期监测矿区地表沉降并掌握其沉降规律势在必行。目前针对矿区地表沉降预测方法较多,但是传统的预测方法存在误差大、收敛速度慢等问题,因此,探索有效的新方法预测矿区地表沉降规律具有重大意义。本文在对矿区地表沉降预测和支持向量机研究现状的基础之上,利用蚁群算法对支持向量机进行优化,并应用于矿区地表沉降预测研究,首先介绍了蚁群算法和支持向量机的基本理论、数学模型及其特点;其次采用改进网格化的蚁群算法优化选择支持向量机参数,基于Visual Studio编程环境,利用C#语言编程实现;最后,通过以某矿区地表沉降监测数据为数据源,采用Fibonacci加权处理方法对原始监测数据进行处理,建立基于蚁群算法的支持向量机矿区地表沉降预测模型,并与传统支持向量机模型预测结果进行对比分析。结果显示,基于蚁群算法的支持向量机预测模型12号、22号、24号点预测值残差分别为0.0015~0.0028m、0.0015~0.0025m、0.0018~0.0028m,传统支持向量机预测模型12号、22号、24号点预测值残差分别为0.0055~0.0086m、0.0060~0.0078m、0.0058~0.0081m。研究结果表明,基于蚁群算法的支持向量机预测模型精度相对较高,优于传统支持向量机模型预测精度,基于蚁群算法的支持向量机预测模型预测值变化曲线更接近实测值,由此我们可以得到基于蚁群算法的支持向量机矿区地表沉降预测模型方法是可行的,在实际工程应用中具有推广价值。
[Abstract]:The subsidence of mining area is a slow and long-term process. With the accumulation of settlement, the total amount of subsidence exceeds the maximum threshold value of the surface, and landslide and other major disasters will occur, in order to avoid the occurrence of disasters. It is imperative to monitor the ground subsidence in mining area for a long time and master its settlement law. At present, there are many forecasting methods for surface subsidence in mining area, but the traditional forecasting methods have many problems, such as large error and slow convergence rate. It is of great significance to explore an effective new method for predicting the rule of ground subsidence in mining area. Based on the research status of ground subsidence prediction and support vector machine in mining area, the ant colony algorithm is used to optimize the support vector machine. The paper introduces the basic theory of ant colony algorithm and support vector machine, mathematical model and its characteristics, and then uses improved mesh ant colony algorithm to optimize the selection of support vector machine parameters. Based on Visual Studio programming environment and C # programming language, finally, the original monitoring data is processed by using Fibonacci weighted processing method, taking the ground subsidence monitoring data of a mining area as the data source. The prediction model of ground subsidence in mining area based on support vector machine (SVM) is established and compared with the prediction results of traditional SVM model. The results show that: 1. The residual error of forecasting model 12, 22 and 24 of support vector machine based on ant colony algorithm is 0.0015 ~ 0. 0028 mN, 0. 0015 ~ 0. 0025 mL, 0. 0018 18 + 0. 0028 m, respectively, and the residual value of point 12, 22 and 24 of traditional SVM prediction model is 0. 0055 ~ 0. 0086 MN, 0. 0060, 0. 0078, 0. 0058, 0. 0081 m. respectively. The results show that, The prediction accuracy of support vector machine based on ant colony algorithm is higher than that of traditional support vector machine, and the prediction curve of support vector machine based on ant colony algorithm is closer to the measured value. From this, we can get the prediction model of ground subsidence based on ant colony algorithm, which is feasible and worth popularizing in practical engineering.
【学位授予单位】：江西理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TD327

【相似文献】